The Global Burden of Meningitis in Children: Challenges withInterpreting Global Health Estimates
Claire Wright 1,* , Natacha Blake 1, Linda Glennie 1, Vinny Smith 1, Rose Bender 2, Hmwe Kyu 2,Han Yong Wunrow 2, Li Liu 3, Diana Yeung 4, Maria Deloria Knoll 5, Brian Wahl 5, James M. Stuart 6,7
3 Department of Population, Family and Reproductive Health and Institute for International Programmes,Johns Hopkins Bloomberg School of Public Health, Baltimore, MD 21205, USA; [email protected]
4 Institute for International Programmes, Department of International Health, Johns Hopkins BloombergSchool of Public Health, Baltimore, MD 21205, USA; [email protected]
5 International Vaccine Access Center, Johns Hopkins Bloomberg School of Public Health,Baltimore, MD 21231, USA; [email protected] (M.D.K.); [email protected] (B.W.)
6 Population Health Sciences, Bristol Medical School, University of Bristol, Bristol BS8 1QU, UK;[email protected]
7 World Health Organization, 1211 Geneva, Switzerland8 Disease Dynamics Unit, Department of Veterinary Medicine, University of Cambridge,
Abstract: The World Health Organization (WHO) has developed a global roadmap to defeat menin-gitis by 2030. To advocate for and track progress of the roadmap, the burden of meningitis as asyndrome and by pathogen must be accurately defined. Three major global health models estimatingmeningitis mortality as a syndrome and/or by causative pathogen were identified and comparedfor the baseline year 2015. Two models, (1) the WHO and the Johns Hopkins Bloomberg Schoolof Public Health’s Maternal and Child Epidemiology Estimation (MCEE) group’s Child MortalityEstimation (WHO-MCEE) and (2) the Institute for Health Metrics and Evaluation (IHME) GlobalBurden of Disease Study (GBD 2017), identified meningitis, encephalitis and neonatal sepsis, collec-tively, to be the second and third largest infectious killers of children under five years, respectively.Global meningitis/encephalitis and neonatal sepsis mortality estimates differed more substantiallybetween models than mortality estimates for selected infectious causes of death and all causes ofdeath combined. Estimates at national level and by pathogen also differed markedly between models.Aligning modelled estimates with additional data sources, such as national or sentinel surveillance,could more accurately define the global burden of meningitis and help track progress against theWHO roadmap.
Keywords: meningitis; child mortality; neonatal sepsis; global health; global health estimates;modelling; Streptococcus pneumoniae; Haemophilus influenzae; Neisseria meningitidis
1. Introduction
The world saw great progress in reducing child mortality over the lifetime of theUnited Nations (UN) Millennium Development Goals (MDGs) with an estimated 54%decline in children under five years of age from 93 deaths per 1000 live births in 1990 to 43per 1000 live births in 2015 [1]. The successor UN Sustainable Development Goals (SDGs)are more ambitious again, and urge that by 2030 we should “end preventable deaths ofnewborns and children under five years of age, with all countries aiming to reduce neonatal
mortality to at least as low as 12 per 1000 live births and under-five mortality to at least aslow as 25 per 1000 live births.” However, with the majority of an estimated 38 deaths per1000 live births in 2019 being caused by preventable and treatable diseases [1], we are along way from achieving this target.
Among these preventable diseases, meningitis has one of the highest fatality rates andthe potential to cause devastating epidemics. Since the turn of the century, we have seenadvances as a result of widespread global introduction of Haemophilus influenzae type b(Hib) and pneumococcal vaccines as well as the roll out of the meningococcal A vaccine,MenAfriVac, across some of the highest incidence areas of sub-Saharan Africa. Despite this,recent estimates have identified that the global burden of meningitis in all age groupsremains high and progress lags substantially behind that of other vaccine preventablediseases [2]. Whilst deaths from measles and tetanus in children under five years areestimated to have decreased by 86% and 92% respectively, between 1990 and 2017, over thesame time period deaths from meningitis are estimated to have decreased by just 51% [3].Despite its burden, meningitis is seldom, if at all, mentioned in key global and regionalhealth documents [4–9].
In response to calls from governments, global health organisations, civil society, publichealth bodies, academia and the private sector, a World Health Organization (WHO)-led collaboration is developing a Defeating Meningitis by 2030 Global Roadmap [10].The Roadmap focuses on the four leading global causes of bacterial meningitis; Neisse-ria meningitidis (meningococcus), Streptococcus pneumoniae (pneumococcus), Haemophilusinfluenzae (Hi), and Streptococcus agalactiae (group B streptococcus (GBS)).
To advocate for a global roadmap to defeat meningitis, the global burden of meningitisas a syndrome in relation to other infectious causes of death needs to be accurately de-scribed, and countries with the highest burden identified, so that efforts and resources canbe targeted effectively. Estimates of pathogen-specific meningitis incidence and mortalityat the global level can identify the need for new vaccines or support wider access to existingones. Tracking trends in pathogen-specific meningitis and syndromic disease over time atthe national and international level is vital to assess the impact of interventions such asvaccines implemented as part of the global roadmap to defeat meningitis.
Vital registration systems and disease surveillance platforms are limited across manygeographies and regions, so there is a reliance on modelled estimates to get a completeglobal picture of disease across all settings but cause of death estimates have been foundto differ across these different modelling efforts [11]. Modelled estimates also attempt toaccount for changes in causes of death over time, but to do so accurately they must beinformed by reliable data to make accurate predictions where real data is lacking.
In this paper we aim to compare the available modelled estimates for cases and deathsfrom meningitis as a syndrome, by causative pathogen and the methods used, in order toassess whether these models can be used with confidence by decision makers to prioritiserecommendations from a plan to defeat meningitis, and by those needing to track progresson the WHO Defeating Meningitis by 2030 Global Roadmap’.
2. Materials and Methods2.1. Identification of Data Sources
Through attending key stakeholder meetings, we identified three modelling effortsthat estimate the global burden of meningitis and neonatal sepsis: (1) WHO and theJohns Hopkins Bloomberg School of Public Health’s Maternal and Child EpidemiologyEstimation (MCEE) group’s Child Mortality Estimation (WHO-MCEE), which estimates15causes of death for children under five years of age [12]; (2) The Institute for Health Metricsand Evaluation (IHME) Global Burden of Disease Study (GBD 2017) which estimates agespecific mortality for 282 causes of death in all ages [3]; and (3) The WHO’s Global HealthEstimates (WHO GHE) which estimates age specific mortality for 136 causes of death inall ages [13].
Microorganisms 2021, 9, 377 3 of 16
Two additional models were also identified that estimated disease burden causedby pathogens of particular relevance to the WHO Defeating Meningitis by 2030 GlobalRoadmap: (1) the WHO-MCEE group’s estimates of the burden of pneumococcal andHib disease in children [14], and (2) the London School of Hygiene and Tropical Medicine(LSHTM) Burden of Group B Streptococcus Worldwide for Pregnant Women, Stillbirths,and Children [15].
Not all of these efforts were directly comparable because they did not provide thesame level of data or use the same indicators of burden (Table 1).
Table 1. Models estimating the global burden of meningitis and neonatal sepsis.
GBD 2017 WHO GHE WHO-MCEESyndromic Model
WHO-MCEEPathogen Model LSHTM
Years 1990–2017 2000–2016 2000–2017 2000–2015 2015
Number ofcountries &territories
195 183 194 194 195
Global under fivepopulation
estimate in 2015678,053,340 673,253,870 ** 671,355,776 ** 657,127,399 *** N/A
Age range
All ages(Including:
Early neonatal: 0–6days
Late neonatal: 7–27days
Post neonatal:28–364 days1–4 years)
All ages(including:0–28 days
1–59 months)
0–59 months(including:0–28 days
1–59 months)
1–59 months 0–89 days
Relevant diseasecategories
Meningitis,neonatal sepsis
and other neonatalinfections
Meningitis *,neonatal sepsisand infections
Meningitis/encephalitis,sepsis and other
infectiousconditions of the
newborn
Meningitis,Non-
pneumonia/non-meningitis (whichis primarily but notexclusively sepsis)
Meningitis,Sepsis
Outputs
Cases,Incidence rate,
Prevalence,Deaths,
Mortality rate,DALYs
Deaths,Mortality rate,
DALYs
Deaths,Mortality rate
Cases,Incidence rate,
Deaths,Mortality rate
Cases,Incidence rate,
Deaths,Mortality rate
Published rate perpopulation
Per 100,000population
Per 100,000population Per 1000 livebirths Per 100,000
population Per 1000 livebirths
Aetiology
Nm,Spn,Hib,
Other
No breakdown byaetiology
No breakdown byaetiology
Nonepidemicdisease from:
Spn,Hib,Nm
GBS
DALYs = Disability Adjusted Life Years; GBS = Group B streptococcus; Hib = Haemophilus influenzae type b; Nm = Neisseria meningitidis(meningococcus); Spn = Streptococcus pneumoniae (pneumococcus).* WHO GHE use a ratio of meningitis to encephalitis deaths obtainedfrom IHME data to separate out MCEE under-five meningitis/encephalitis estimates. ** Estimates derived from UN World PopulationProspects 2017. Differences between WHO GHE and WHO-MCEE population estimates likely due to draft estimates circulating prior tofinal publication. *** Derived from UN World Population Prospects 2015
As WHO GHE estimates were an amalgamation of historical models (WHO-MCEE’s2000–2016 and IHME’s GBD 2016) we did not consider them further in our analysis. We didnot include GBS estimates from LSHTM in our analysis because the age categorisation
Microorganisms 2021, 9, 377 4 of 16
(0–89 days) did not correspond with the disaggregated age categories of the other modelsand so did not allow for meaningful comparison.
2.2. Analysis of Data Sources
The scale of the global burden of meningitis deaths relative to all causes and leadinginfectious causes of death was assessed by comparing, death and mortality estimates fromGBD 2017 and the WHO-MCEE’s 2000–2017 model according to the following syndromiccause of death categories “All causes”, “Infectious disease”, “meningitis/encephalitis” and“neonatal sepsis”.
We considered the burden of meningitis and neonatal sepsis together for the purposesof comparison with other leading infectious causes of death because distinguishing betweenthese syndromes is almost impossible based on clinical signs alone in the neonate [16,17].Lumbar puncture (LP) and analysis of the cerebrospinal fluid is the only reliable way ofconfirming a case of meningitis. However, in many countries there is a shortage of trainedstaff to perform LP [18], and in low-income settings as few as 2% of neonates with infectionmight have an LP or blood sample taken [19].
The WHO-MCEE have historically estimated sepsis and meningitis in the neonatalperiod within the same cause category because of difficulties in distinguishing betweenthese clinical syndromes in this age group. These causes were estimated separately forthe first time in their latest modelling round by using the ratio of neonatal meningitisand neonatal sepsis deaths derived from IHME estimates. Because WHO-MCEE estimatemeningitis/encephalitis as one cause category, GBD 2017 meningitis and encephalitisdeaths were amalgamated for the purpose of comparison.
Denominators used to report mortality rates were standardised across the models and,where necessary, recalculated to be expressed as deaths per 1000 live births in the neonatalperiod and deaths per 100,000 population in the post neonatal period. GBD 2017 mortalityrates in the neonatal period were calculated from IHME live birth estimates for the year2015. WHO-MCEE postneonatal mortality rates were calculated using UN populationestimates for the year 2015 [20].
Priority geographical areas for targeting a plan to defeat meningitis were identifiedfrom country-specific GBD 2017 and WHO-MCEE meningitis/encephalitis mortality esti-mates for the year 2015 in children under five years.
Meningitis mortality and incidence estimates according to pathogen over time (2000–2015) were analysed using estimates produced by GBD 2017 and the WHO-MCEE pathogenmodel. Meningococcal meningitis is commonly associated with epidemics. As WHO-MCEE meningococcal meningitis estimates did not account for deaths and cases resultingfrom epidemics, estimates for ‘Hib meningitis’ and ‘pneumococcal meningitis’ mortalityand incidence in the post neonatal period (28 days–<5 years) were the categories and agegroup used for comparison.
An analysis of the estimation methodology for each model was also undertaken in anattempt to explain any inconsistencies between models.
3. Results3.1. Global Meningitis and Neonatal Sepsis Mortality Estimates in Children Aged Under FiveYears
Overall, the WHO-MCEE estimated there to be approximately 100,000 fewer deaths inthe under-five age group than the GBD 2017, with proportionally more under-five deathsoccurring in the neonatal period (46% compared to GBD 2017’s 43%).
The GBD 2017 estimated 34% more deaths from meningitis/encephalitis than theWHO-MCEE in the year 2015 (190,515 and 142,841, respectively) (Table 2). Meningitismade up the majority of the GBD 2017 combined meningitis/encephalitis category; 87% inunder five-year-olds, 86% in 1–59 months and 93% in 0–28 days.
Microorganisms 2021, 9, 377 5 of 16
Table 2. Estimated deaths by cause and model for the year 2015 in children under five years of age.
GBD 2017 WHO-MCEE Pathogen Model Difference *
n Rate † n Rate † %
All causes
Under 55,917,285 872.69 5,792,509
862.81 a 2%(5,723,776–6,120,099) (844.15–902.60) (5,573,633–
6,123,477)
1–59 months3,354,404 502.51 3,122,698
473.02 a 7%(3,231,491–3,483,015) (484.10–521.78) (2,700,899–
3,581,030
0–28 days 2,562,881 18.40 2,669,811 19.01 −4%(2,478,272–2,655,261) (17.20–19.58) (2,542,447–
2,872,734) (18.10–20.50)
Infectiousdiseases **
Under 52,519,567 371.59 2,426,882
361.49 a 4%(2,379,024–2,671,856) (350.86–394.05) (2,279,602–
3,169,783)
1–59 months1,967,826 294.79 1,810,771
274.29 a 8%(1,847,763–2,091,762) (276.81–313.36) (1,703,587–
2,350,572)
0–28 days 551,740 3.96 616,111 4.39 −11%(510,918–603,527 (3.60–4.38) (605,290–
877,610) (4.31–6.25)
Meningitis &Encephalitis
Under 5190,515 28.10 142,841
21.28 a 29%(163,374–217,259) (24.09–32.04) (87,427–
178,552)
1–59 months167,880 25.15 105,406
15.97 a 46%(143,529–192,447) (21.50–28.83) (87,188–
145,213)
0–28 days 22,636 0.16 37,435 0.27 −49%(18,532–25,642) (0.13–0.19) (157–51,299) (0.001–0.37)
Neonatal sepsis
Under 5211,273 31.16 364,188
54.25 a −53%(186,657–275,821) (27.53–40.68) (282,744–
524,021)
1–59 months12,693 1.90 386 b
0.06 a 188%(10,626–16,586) (1.59–2.48) (14–579)
0–28 days 198,580 1.43 363,802 2.59 −59%(175,866–263,096) (1.24–1.86) (282,341–
523,853) (2.01–3.73)
* Percent difference (n) = (GBD 2017–WHO-MCEE)/((GBD 2017 + WHO-MCEE)/2) × 100. ** Sum of specific infectious diseases fromWHO-MCEE cause list (HIV/AIDS; diarrhoeal diseases; tetanus; measles; meningitis/encephalitis; malaria; acute respiratory infections;sepsis and other infectious conditions of the newborn). † Rates per 100,000 population in ‘Under 5′ and ‘1–59 months’, and per 1000livebirths for ‘0–28 days’. a Uncertainty intervals not available–rate calculated using n and under-5 population statistic from UN WPP 2017Revision–year 2015 (1–59 months calculated using 59/60 months population). b Figures only account for neonatal sepsis deaths in China.
However, the WHO-MCEE estimated >100,000 more deaths than the GBD 2017 whenneonatal sepsis deaths were combined with meningitis/encephalitis, due to the WHO-MCEE’s much higher estimate of neonatal sepsis deaths. Uncertainty intervals do notoverlap between modelled estimates of deaths from neonatal sepsis in any of the agecategories. This section may be divided by subheadings. It should provide a conciseand precise description of the experimental results, their interpretation, as well as theexperimental conclusions that can be drawn.
The WHO-MCEE model estimated meningitis/encephalitis and neonatal sepsis as thesecond largest infectious cause of death, co-ranked with diarrhoeal diseases, in childrenaged under five years in 2015, after acute respiratory infections (Figure 1). In contrast theGBD 2017 estimated this cause category to be the third largest infectious case of death afteracute respiratory infections and diarrhoeal diseases.
Microorganisms 2021, 9, 377 6 of 16Microorganisms 2021, 9, x FOR PEER REVIEW 6 of 17
Figure 1. Meningitis/encephalitis and neonatal sepsis mortality burden estimates by model in rela-tion to other selected infectious causes of death in children under five for the year 2015.
At the country level, there was considerable variability in estimates of burden per population. For meningitis/encephalitis mortality rates, the WHO-MCEE model ranked Somalia highest for the year 2015 (139.7 deaths per 100,000 population), whilst the GBD 2017 ranked Somalia 11th highest for the same year (68.6 deaths per 100,000).
For numbers of deaths, both models attribute approximately 70% of all meningi-tis/encephalitis deaths in children under five years to just 12 countries including India, Nigeria, Pakistan, Democratic Republic of Congo (DRC), Ethiopia, Niger, Afghanistan, Mali, Uganda and China. However, whilst Somalia and Chad feature in the top 12 (ranked 7th and 8th respectively) in the WHO-MCEE estimates, they did not feature in the GBD 2017 top 12, where Indonesia and Burkina Faso featured instead (ranked 8th and 9th high-est respectively) (Figure 2).
Figure 1. Meningitis/encephalitis and neonatal sepsis mortality burden estimates by model in relation to other selectedinfectious causes of death in children under five for the year 2015.
At the country level, there was considerable variability in estimates of burden perpopulation. For meningitis/encephalitis mortality rates, the WHO-MCEE model rankedSomalia highest for the year 2015 (139.7 deaths per 100,000 population), whilst the GBD2017 ranked Somalia 11th highest for the same year (68.6 deaths per 100,000).
For numbers of deaths, both models attribute approximately 70% of all meningi-tis/encephalitis deaths in children under five years to just 12 countries including India,Nigeria, Pakistan, Democratic Republic of Congo (DRC), Ethiopia, Niger, Afghanistan,Mali, Uganda and China. However, whilst Somalia and Chad feature in the top 12 (ranked7th and 8th respectively) in the WHO-MCEE estimates, they did not feature in the GBD2017 top 12, where Indonesia and Burkina Faso featured instead (ranked 8th and 9thhighest respectively) (Figure 2).
3.2. Meningitis Incidence and Mortality Estimates by Aetiology in Children Aged UnderFive Years
The GBD 2017 and WHO-MCEE’s pathogen models both estimated pneumococcaland Hib meningitis mortality and incidence in children aged 1 to 59 months at the nationaland global levels.
A comparison of the global estimates for the year 2015 (Table 3) showed that bothmodels agree there were more cases of pneumococcal meningitis than Hib meningitis in2015. However, whilst the GBD 2017 estimated around twice as many deaths from Hibmeningitis compared to pneumococcal meningitis, the WHO-MCEE estimated aroundfive times more deaths from pneumococcal meningitis than from Hib meningitis in thesame year.
Microorganisms 2021, 9, 377 7 of 16Microorganisms 2021, 9, x FOR PEER REVIEW 7 of 17
Figure 2. Top twelve ranking countries by meningitis/encephalitis mortality (number and rate) according to model for the year 2015.
3.2. Meningitis Incidence and Mortality Estimates by Aetiology in Children Aged Under Five Years
The GBD 2017 and WHO-MCEE’s pathogen models both estimated pneumococcal and Hib meningitis mortality and incidence in children aged 1 to 59 months at the national and global levels.
A comparison of the global estimates for the year 2015 (Table 3) showed that both models agree there were more cases of pneumococcal meningitis than Hib meningitis in 2015. However, whilst the GBD 2017 estimated around twice as many deaths from Hib meningitis compared to pneumococcal meningitis, the WHO-MCEE estimated around five times more deaths from pneumococcal meningitis than from Hib meningitis in the same year.
Figure 2. Top twelve ranking countries by meningitis/encephalitis mortality (number and rate) according to model for theyear 2015.
Table 3. Global aetiology-specific meningitis deaths and cases, 2015, in children aged 1–59 months.
* Percent difference = (GBD 2017–WHO-MCEE)/((GBD 2017 + WHO-MCEE)/2) × 100. † Rates per 100,000 population
Microorganisms 2021, 9, 377 8 of 16
Despite major differences in the relative proportions of meningitis deaths attributedto Hib and pneumococcal bacteria between models, both models agreed that Hib andpneumococcal meningitis combined were the underlying cause of approximately 40% ofall meningitis/encephalitis deaths globally.
When comparisons were made between the modelled estimates for Hib and pneu-mococcal meningitis incidence and mortality over time (Figure 3), both models showeda steeper decline in Hib meningitis incidence and mortality compared to pneumococcalmeningitis mortality, which is consistent with wider roll-out of Hib vaccination globallycompared to pneumococcal vaccination. However, the GBD 2017 consistently reportedmuch higher incidence of pneumococcal and Hib meningitis over time compared to theWHO-MCEE. The GBD 2017 estimated Hib and pneumococcal incidence to be 31 and 40cases per 100,000, respectively, in 2015 compared to the WHO-MCEE estimates of aroundfive and 13 cases per 100,000 for Hib and pneumococcal meningitis, respectively.
Microorganisms 2021, 9, x FOR PEER REVIEW 8 of 17
Table 3. Global aetiology-specific meningitis deaths and cases, 2015, in children aged 1–59 months.
Despite major differences in the relative proportions of meningitis deaths attributed to Hib and pneumococcal bacteria between models, both models agreed that Hib and pneumococcal meningitis combined were the underlying cause of approximately 40% of all meningitis/encephalitis deaths globally.
When comparisons were made between the modelled estimates for Hib and pneu-mococcal meningitis incidence and mortality over time (Figure 3), both models showed a steeper decline in Hib meningitis incidence and mortality compared to pneumococcal meningitis mortality, which is consistent with wider roll-out of Hib vaccination globally compared to pneumococcal vaccination. However, the GBD 2017 consistently reported much higher incidence of pneumococcal and Hib meningitis over time compared to the WHO-MCEE. The GBD 2017 estimated Hib and pneumococcal incidence to be 31 and 40 cases per 100,000, respectively, in 2015 compared to the WHO-MCEE estimates of around five and 13 cases per 100,000 for Hib and pneumococcal meningitis, respectively.
Figure 3. Estimated Hib/pneumococcal meningitis mortality and incidence amongst children aged 1–59 months accordingto model in relation to the proportion of children unimmunised with Hib vaccine and pneumococcal conjugate vaccine(PCV) over time.
Of note is that case fatality rates (CFRs) differed dramatically between the two setsof estimates. CFRs derived from WHO-MCEE global cases, and deaths estimates for Hiband pneumococcal meningitis in 2015, were 23% and 45%, respectively. However, CFRscalculated from GBD 2017 estimates were 8% for pneumococcal meningitis, 19% for Hibmeningitis and 8% for meningococcal meningitis. Evidence from the literature closelyagrees with the WHO-MCEE CFRs, consistently reporting higher CFRs from pneumococcalmeningitis compared to Hib meningitis and meningococcal meningitis [21–26].
Microorganisms 2021, 9, 377 9 of 16
3.3. Modelling Methodology Which Could Account for Differences in Mortality and PathogenSpecific Estimates
Figure 4 depicts a simplified methodology for both modelling approaches. A moredetailed explanation is provided in the appendix, and full methodological approachesare also outlined elsewhere [3,27]. When calculating the meningitis death envelope, bothmodels used country-specific death data from vital registration and other sources andapplied statistical modelling to fill gaps in the data using country-specific covariates anddrawing on trends observed where data was more complete. Whilst the GBD includedintervention covariates (such as vaccine coverage) within their cause of death ensemblemodelling (CODEm) (Figure 4), the WHO-MCEE model used intervention covariates inboth their modelling, and also in post hoc adjustments, to redistribute causes accountingfor interventions. Details of the covariates used by the models are available in the Sup-plementary Materials. Both models ensured that the sum of deaths attributed to differentcauses fitted within a total all-cause mortality envelope calculated from surveys, censusesand vital registration data.
Whilst there was little difference between estimated mortality from all causes andinfectious diseases in children under five years (2% and 4% difference in estimateddeaths, respectively), between models there was a marked difference between menin-gitis/encephalitis and neonatal sepsis mortality estimates in this age group (29 and 53percent difference, respectively) (Table 2).
Further investigation into the modelling methods and underlying data showed thatcountries with the highest meningitis burden have the lowest quality death registrationdata. Whilst this is also the case for all causes of death, a higher proportion of meningi-tis/encephalitis death estimates were based on extrapolating from low-quality underlyingdata compared to all-cause death estimates. For example, 77% of meningitis/encephalitisdeaths came from countries with no or very low-quality death registration data (scaled 0 to1) compared to 60% of deaths due to all causes in the GBD 2017 model. Likewise, in theWHO-MCEE model, 95% of meningitis/encephalitis deaths were estimated using mod-elling underpinned by verbal autopsy (VA) studies compared to 90% of all cause deathsdue to these countries having poor quality death registration data (see SupplementaryMaterials). As would be expected, there were greater differences between estimates fromcountries with low-quality underlying data compared to those with higher quality data(Figure 5).
Microorganisms 2021, 9, x FOR PEER REVIEW 9 of 17
Figure 3. Estimated Hib/pneumococcal meningitis mortality and incidence amongst children aged 1–59 months according to model in relation to the proportion of children unimmunised with Hib vaccine and pneumococcal conjugate vaccine (PCV) over time.
Of note is that case fatality rates (CFRs) differed dramatically between the two sets of estimates. CFRs derived from WHO-MCEE global cases, and deaths estimates for Hib and pneumococcal meningitis in 2015, were 23% and 45%, respectively. However, CFRs calculated from GBD 2017 estimates were 8% for pneumococcal meningitis, 19% for Hib meningitis and 8% for meningococcal meningitis. Evidence from the literature closely agrees with the WHO-MCEE CFRs, consistently reporting higher CFRs from pneumococ-cal meningitis compared to Hib meningitis and meningococcal meningitis [21–26].
3.3. Modelling Methodology Which Could Account for Differences in Mortality and Pathogen Specific Estimates
Figure 4 depicts a simplified methodology for both modelling approaches. A more detailed explanation is provided in the appendix, and full methodological approaches are also outlined elsewhere [3,27]. When calculating the meningitis death envelope, both models used country-specific death data from vital registration and other sources and ap-plied statistical modelling to fill gaps in the data using country-specific covariates and drawing on trends observed where data was more complete. Whilst the GBD included intervention covariates (such as vaccine coverage) within their cause of death ensemble modelling (CODEm) (Figure 4), the WHO-MCEE model used intervention covariates in both their modelling, and also in post hoc adjustments, to redistribute causes accounting for interventions. Details of the covariates used by the models are available in the Supple-mentary Materials. Both models ensured that the sum of deaths attributed to different causes fitted within a total all-cause mortality envelope calculated from surveys, censuses and vital registration data.
Figure 4. Cont.
Microorganisms 2021, 9, 377 10 of 16Microorganisms 2021, 9, x FOR PEER REVIEW 10 of 17
Figure 4. Simplified schematic of the different mortality modelling approaches. VR Data—Data from 76 countries with high quality VR data covering >80% of the population was mapped di-rectly to cause of death categories (see appendix for ICD10 codes mapped to meningitis and sepsis and other severe infections in the neonatal period). VRMCM—Data from the countries with high quality VR data was used to fit a multinomial logistic regression model which was used to predict cause of death proportions in 38 low mortality countries (<35 deaths/1000 live births 2000–2010) with low quality VR data. Covariates used in the model are provided in appendix. VAMCM—In 78 high mortality countries (>35 deaths/1000 live births 2000–2010) verbal autopsy data from 119 research studies in 39 high mortality countries was used to fit a multinomial model to predict causes of death. Cause of death proportions for India were estimated using a combination of VAMCM for the neonatal period and data from the million deaths study and INDEPTH sites in India for the post neonatal period. See appendix for model covariates Other – Cause of death pro-portions for China were estimated using data from the China Maternal and Child Health Surveil-lance system. A complete explanation of methods used to produce WHO/MCEE estimates is out-lined elsewhere [27].
Whilst there was little difference between estimated mortality from all causes and infectious diseases in children under five years (2% and 4% difference in estimated deaths, respectively), between models there was a marked difference between meningitis/enceph-alitis and neonatal sepsis mortality estimates in this age group (29 and 53 percent differ-ence, respectively) (Table 2).
Further investigation into the modelling methods and underlying data showed that countries with the highest meningitis burden have the lowest quality death registration data. Whilst this is also the case for all causes of death, a higher proportion of meningi-tis/encephalitis death estimates were based on extrapolating from low-quality underlying data compared to all-cause death estimates. For example, 77% of meningitis/encephalitis deaths came from countries with no or very low-quality death registration data (scaled 0 to 1) compared to 60% of deaths due to all causes in the GBD 2017 model. Likewise, in the WHO-MCEE model, 95% of meningitis/encephalitis deaths were estimated using model-ling underpinned by verbal autopsy (VA) studies compared to 90% of all cause deaths due to these countries having poor quality death registration data (see Supplementary Materials). As would be expected, there were greater differences between estimates from countries with low-quality underlying data compared to those with higher quality data (Figure 5).
Figure 4. Simplified schematic of the different mortality modelling approaches. VR Data—Data from 76 countries withhigh quality VR data covering >80% of the population was mapped directly to cause of death categories (see appendix forICD10 codes mapped to meningitis and sepsis and other severe infections in the neonatal period). VRMCM—Data fromthe countries with high quality VR data was used to fit a multinomial logistic regression model which was used to predictcause of death proportions in 38 low mortality countries (<35 deaths/1000 live births 2000–2010) with low quality VR data.Covariates used in the model are provided in appendix. VAMCM—In 78 high mortality countries (>35 deaths/1000 livebirths 2000–2010) verbal autopsy data from 119 research studies in 39 high mortality countries was used to fit a multinomialmodel to predict causes of death. Cause of death proportions for India were estimated using a combination of VAMCMfor the neonatal period and data from the million deaths study and INDEPTH sites in India for the post neonatal period.See appendix for model covariates Other – Cause of death proportions for China were estimated using data from theChina Maternal and Child Health Surveillance system. A complete explanation of methods used to produce WHO/MCEEestimates is outlined elsewhere [27].
To estimate meningitis mortality by aetiology, both models applied a proportional splitby pathogen to the country-specific meningitis death envelope and adjusted for vaccinecoverage. Whilst GBD 2017 pathogen specific mortality proportions were informed by vitalregistration (VR) data from data rich locations, the WHO-MCEE model based mortalityproportions on studies reporting the distribution of pathogen-specific meningitis cases ad-justed by pathogen-specific CFRs to derive proportions of deaths. This approach was useddue to a lack of literature reporting meningitis mortality fractions by pathogen. To adjustpathogen-specific estimates according to vaccine coverage, IHME ran a metaregressionmodel (DisMod-MR 2.1) with pneumococcal and Hib vaccine coverage as covariates driv-ing down the proportions of disease attributed to those pathogens. The WHO-MCEE modelused a deterministic approach to account for vaccine use by calculating the percentagereduction in disease as a result of vaccine efficacy, coverage and, in the case of PCV, thevaccine product and proportion of disease caused by vaccine-specific serotypes.
The models used very different approaches for estimating incidence by aetiology.The GBD 2017 calculated meningitis incidence independently from meningitis mortal-ity using incidence data gathered from hospital records, claims data and a systematicreview of the literature capturing incidence studies. The WHO-MCEE incidence esti-mates were derived by dividing pathogen-specific death estimates by literature-derivedCFRs. The WHO-MCEE also published an update to a previous incidence-based modelfor Hib and pneumococcal meningitis [28], which predicted even lower incidence rates forpneumococcal meningitis and similar rates for Hib.
Microorganisms 2021, 9, 377 11 of 16Microorganisms 2021, 9, x FOR PEER REVIEW 11 of 17
Figure 5. Absolute difference between WHO-MCEE and GBD 2017 meningitis/encephalitis mortal-ity estimates according to country.
To estimate meningitis mortality by aetiology, both models applied a proportional split by pathogen to the country-specific meningitis death envelope and adjusted for vac-cine coverage. Whilst GBD 2017 pathogen specific mortality proportions were informed by vital registration (VR) data from data rich locations, the WHO-MCEE model based mortality proportions on studies reporting the distribution of pathogen-specific meningi-tis cases adjusted by pathogen-specific CFRs to derive proportions of deaths. This ap-proach was used due to a lack of literature reporting meningitis mortality fractions by pathogen. To adjust pathogen-specific estimates according to vaccine coverage, IHME ran
Figure 5. Absolute difference between WHO-MCEE and GBD 2017 meningitis/encephalitis mortality estimates accordingto country.
4. Discussion
Despite major differences in the number of deaths attributed to meningitis, bothmodels agree that there is a substantial burden of disease, with meningitis as either the2nd or 3rd most important infectious syndrome. By far the biggest burden of meningitis isestimated to occur in countries with low quality or no death registration data where thesemodels rely heavily on extrapolating from VA studies. Accurately attributing meningitis asa cause of death using VA is extremely challenging [29–31] and could lead to meningitisas a cause of death being underestimated. VA has a high specificity but low to moderate
Microorganisms 2021, 9, 377 12 of 16
sensitivity for meningitis [32–34] and can easily attribute death from meningitis to adifferent cause, especially in malaria endemic regions where severe febrile illness is oftenassumed to be malaria [35–37].
If these syndromic models systematically underestimate deaths from meningitis, thiswould result in an underestimate of incidence by pathogen in the WHO-MCEE modelbecause incidence is derived by dividing estimated deaths by CFR based on location andpathogen. The GBD 2017 estimated pathogen-specific incidence separately to pathogen-specific deaths and produced higher estimates than the WHO-MCEE model, but theincidence estimates were out of line with deaths when literature-derived CFRs wereapplied. Following a meeting where results from this analysis were presented to allmodelling groups, the IHME amended their methodology for calculating pathogen-specificincidence. In the recently published GBD 2019 model [38], published studies and hospitaldata were used to estimate pathogen-specific CFRs as a function of healthcare access andquality. Pathogen specific mortality was then derived from estimates of pathogen-specificincidence and CFRs.
Using global health estimates to derive baseline numbers and targets against whichprogress can be measured is challenging. Estimates for the entire time series are updatedwith successive model iterations as new input data are considered and amendments aremade to statistical modelling processes. This means that baseline estimates for a given yearfluctuate with successive model iterations.
It is vital that the methods used to derive estimates are clearly communicated. Acrossmodels it was unclear from published methods exactly how neonatal meningitis as a causewas disaggregated from neonatal sepsis, and other infectious conditions of the newborn,when we know that the majority of the underlying input data does not distinguish betweenthese two causes of death. Unless methods are made transparent, it is difficult for policymakers to understand, and therefore trust, model outputs [39].
Experts responsible for monitoring progress also need to know exactly how estimateswere derived in order to assess whether they are capable of measuring progress againstcertain indicators. Whilst both models accounted for PCV and Hib vaccine impact, theydid so using substantially different methods. The IHME’s GBD 2017 study accountedfor vaccine impact by finding existing relationships between vaccine coverage and theproportion of pathogen-specific meningitis targeted by the vaccine (from countries wheredata is available) and using these existing relationships to make predictions where datais unavailable. Whilst this approach has an advantage of using as much raw data aspossible, it does not distinguish between differences in vaccine products and the varyingefficacy associated with different dosing schedules between countries. Although incidenceproportion models included data from some countries in sub-Saharan Africa and Asia, theuse of VR data alone to determine proportional cause of death means that vaccine effectson pathogen-specific mortality in high mortality countries with no vital registration dataare heavily reliant on effects demonstrated in data-rich low-mortality countries. The WHO-MCEE, on the other hand, make predictions where data is sparse/unavailable by simulatingthe effect of a given vaccine over time on a country specific basis. Assumptions aboutvaccine impact are transparent and take into account differences in vaccine formulationsand dosing schedules, but they may be applied to a pathogen specific meningitis deathestimate which is highly uncertain.
It is also important for decision makers to be aware that even in data-rich locations,global health estimates for the most recent year can be based on predictions rather thanreal underlying data. These estimates may, therefore, be unsuitable for tracking change asa result of a recent intervention, especially if the intervention has not been accounted for asa covariate in the model.
The IHME’s GBD model is currently the only available complete source of informationabout the global and national burden of meningitis amongst all age groups and for mostof the pathogens of interest to the global roadmap to defeat meningitis. The IHME havealso improved some of their methods for the latest round of estimates by including more
Microorganisms 2021, 9, 377 13 of 16
surveillance data from high mortality settings in the GBD 2019 than was included inthe GBD 2017. Additionally, there are plans for future versions of the IHME’s model toinclude estimates on the incidence and mortality from GBS meningitis, one of the majorcauses of meningitis in neonates worldwide. However, tracking outputs from multiplemodels in parallel has advantages in identifying areas of higher uncertainty, generatingopportunities for modellers to improve methods and prioritising further primary datacollection/strengthening surveillance. An interactive visualisation has been created totrack progress using estimates from all of the major global health estimation models [40].
None of the models we assessed were able to accurately account for the fluctuatingscales of periodic, large epidemics of meningitis, which are irregular and unpredictablein nature. Whilst GBD 2017 attempted to account for epidemic meningococcal meningitisdeaths by adding these to the meningitis death envelope, they did not use equivalentmethodology to account for epidemic meningococcal meningitis cases. The WHO-MCEEsyndromic model attempted to account for epidemic disease by estimating the averageincrease in deaths in epidemic years relative to nonepidemic years and adding these toestimates in years with epidemics identified by WHO surveillance reports and publishedliterature. This increases estimates during an epidemic year, but the underlying datafrom the country are not always reliable, and it does not accurately reflect the variationin the size of the epidemic for a given year. The WHO-MCEE pathogenic model onlyestimated pathogen-specific deaths for endemic disease, removing the simulated effects ofepidemics from the syndromic model before applying proportional splits to the remainingmeningitis envelope. Therefore, neither model estimating pathogen specific causes ofmeningitis was able to account for epidemic pneumococcal meningitis, yet this is animportant consideration because it has been demonstrated as having a significant mortalityburden [41].
Considering the current limitations of modelled meningitis estimates, it is desirableto track progress alongside additional data where possible. Countries across the Africanmeningitis belt experience the highest burden of meningitis globally because they aresusceptible to large and devastating outbreaks of meningococcal disease linked to climaticfactors such as dry winds, low humidity and high levels of dust in the air [42]. Whilstmany of these countries have poor death registration systems, they have relatively richand complementary meningitis surveillance systems. Since 2003 an enhanced meningitissurveillance network has been established across the meningitis belt to strengthen out-break detection and enable a rapid response to outbreaks of meningococcal disease acrossthe region [43]. The network now covers 24 countries, reporting suspected cases anddeaths from meningitis to the WHO intercountry support team (WHO/IST) each weekduring the meningitis season and every month for the rest of the year [44]. Case-basedsurveillance systems have been established in five countries within the region allowing forcomprehensive information on CFRs by age [45].
Triangulating modelled estimates against surveillance data provides the opportunityto reality-check modelled outputs. Utilising surveillance data in combination with evidenceof age and regionally specific CFRs has already successfully been used by experts wishingto monitor global progress towards the 2005 measles mortality reduction goal becausemeasles mortality estimates calculated from vital registration data were considered anunreliable way to track progress [46]. Surveillance data for meningitis is not currentlyavailable for every country worldwide. However, comprehensive roll out of Hib andpneumococcal vaccines is driving down incidence and mortality from meningitis causedby these pathogens across the globe. Improved pathogen-specific surveillance informedby accurate and timely laboratory diagnosis is required to adequately assess the impactof these important life-saving interventions. This is particularly important for countriestransitioning out of Gavi support which need to justify national investments in thesevaccines. Additionally, all member states of the UN have committed to achieving universalhealth coverage by signing up to the SDGs, so there is reason to believe that the availabilityof good quality surveillance data will improve over time as health systems are strengthened.
Microorganisms 2021, 9, 377 14 of 16
More work is required to provide credible meningitis burden estimates for measuringprogress. Currently meningitis mortality estimates are highly uncertain because the modelsrely heavily on death registration data, which is largely missing or incomplete in countrieswith the highest meningitis burden. Additionally, since postmortem examination is rarelyperformed in countries without vital registration systems, and the symptoms of meningitiscan easily be mistaken for other diseases, there is a risk that the mortality burden ofmeningitis could be underestimated. Encouragingly, better data on cause of death arebecoming available in regions where child mortality rates are the highest through the useof minimally invasive tissue sampling [47,48] and inclusion of these data in future modelscould considerably improve the reliability of their outputs.
5. Conclusions
Global meningitis estimates should be interpreted with caution. Tracking progresstowards controlling this disease should also include analysis of real surveillance data whereavailable. The WHO Defeating Meningitis by 2030 Global Roadmap will improve aware-ness, diagnosis and surveillance of meningitis. As the roadmap drives more comprehensivedata on meningitis, a convergence in modelled estimates and a more reliable picture ofreductions in the burden of meningitis are anticipated.
Supplementary Materials: The following are available online at https://www.mdpi.com/2076-2607/9/2/377/s1 [49–51].
Author Contributions: Conceptualization, C.W., N.B., L.G., V.S., J.M.S. and C.T.; methodology, C.W.,N.B., L.G., J.M.S., C.T.; validation, R.B., H.K., H.Y.W., L.L., D.Y., M.D.K., B.W.; formal analysis, C.W.and N.B.; writing—original draft preparation, C.W. and N.B.; writing—review and editing, C.W.,L.G., V.S., R.B., H.K., H.Y.W., L.L., D.Y., M.D.K., B.W., J.M.S., C.T.; visualization, C.W.; supervision,L.G., J.M.S. and C.T. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no specific funding, but Meningitis Research Foundation hasreceived unrestricted educational grants from GSK, Pfizer and Sanofi which allowed this study totake place.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Publicly available datasets were analyzed in this study. GBD 2017 esti-mates are available from http://ghdx.healthdata.org/gbd-2017; WHO-MCEE syndromic meningitisestimates are available from http://158.232.12.119/healthinfo/global_burden_disease/estimates/en/index2.html and WHO-MCEE syndromic estimates are available from their publication [14].
Acknowledgments: This work arose from a meeting to evaluate the global burden of meningitisestimates convened by the Gates Foundation in November 2018.
Conflicts of Interest: C.T. received a consulting payment from GSK in 2018 outside the submittedwork. C.W., N.B., L.G., V.S. took part in this research as employees of Meningitis Research Foundation,which has received unrestricted educational grants from GSK, Pfizer and Sanofi. All other authorsdeclare no conflict of interest. GSK, Pfizer and Sanofi had no role in the design of this study; in thecollection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision topublish the results.
References1. United Nations Inter-agency Group for Child Mortality Estimation (UN IGME). Levels & Trends in Child Mortality: Report
2020, Estimates Developed by the UN Inter-Agency Group for Child Mortality Estimation. 2020. Available online: https://www.un.org/development/desa/pd/news/levels-and-trends-child-mortality-2020-report (accessed on 1 December 2020).
2. Zunt, J.R.; Kassebaum, N.J.; Blake, N.; Glennie, L.; Wright, C.; Nichols, E.; Abd-Allah, F.; Abdela, J.; Abdelalim, A.; A Adamu,A.; et al. Global, regional, and national burden of meningitis, 1990–2016: A systematic analysis for the Global Burden of DiseaseStudy 2016. Lancet Neurol. 2018, 17, 1061–1082. [CrossRef]
3. GBD Causes of Death Collaborators. Global, regional, and national age-sex-specific mortality for 282 causes of death in 195countries and territories, 1980–2017: A systematic analysis for the Global Burden of Disease Study 2017. Lancet 2018, 392,1736–1788. [CrossRef]
4. World Health Organization. Global Action Plan on Antimicrobial Resistance. Available online: https://www.who.int/antimicrobial-resistance/publications/global-action-plan/en/ (accessed on 3 February 2020).
5. Global Strategy for Women’s, Children’s, and Adolescents’ Health 2016–2030. Available online: https://www.who.int/life-course/partners/global-strategy/globalstrategyreport2016-2030-lowres.pdf (accessed on 3 February 2020).
6. World Health Organization/The United Nations Children’s Fund (UNICEF). Global Action Plan for Prevention and Control ofPneumonia (GAPP). Available online: https://www.unicef.org/media/files/GAPP3_web.pdf (accessed on 3 February 2020).
7. World Health Organization/The United Nations Children’s Fund (UNICEF). Ending Preventable Child Deaths from Pneumoniaand Diarrhoea by 2025. The Integrated Global Action Plan for Pneumonia and Diarrhoea (GAPPD). Available online: https://www.unicef.org/media/files/Final_GAPPD_main_Report-_EN-8_April_2013.pdf (accessed on 3 February 2020).
8. Transforming Our World: The 2030 Agenda for Sustainable Development. Goal 3. Ensure Healthy Lives and Promote Well-Beingfor All at All Ages. Available online: https://sustainabledevelopment.un.org/post2015/transformingourworld (accessed on3 February 2020).
9. Working Group of the Specialised Technical Committee on Health, Population and Drug Control, Africa Health Strategy 2016–2030. Available online: https://au.int/sites/default/files/newsevents/workingdocuments/27513-wd-sa16951_e_africa_health_strategy-1.pdf (accessed on 3 February 2020).
10. World Health Organization. Defeating meningitis by 2030. Available online: https://www.who.int/initiatives/defeating-meningitis-by-2030 (accessed on 8 February 2021).
11. Rudan, I.; Campbell, H.; Marusic, A.; Sridhar, D.; Nair, H.; Adeloye, D.; Theodoratou, E.; Chan, K.Y. Assembling GHERG: Could“academic crowd–sourcing” address gaps in global health estimates? J. Glob. Heal. 2015, 5, 010101. [CrossRef]
12. World Health Organization. Disease Burden and Mortality Estimates. Child Causes of Death, 2000–2017. Available online:https://www.who.int/healthinfo/global_burden_disease/estimates/en/index2.html (accessed on 14 October 2019).
13. World Health Organization. Global Health Estimates 2016: Deaths by Cause, Age, Sex, by Country and by Region, 2000–2016.Available online: http://www.who.int/healthinfo/global_burden_disease/estimates/en/ (accessed on 10 June 2019).
14. Wahl, B.; O’Brien, K.L.; Greenbaum, A.; Majumder, A.; Liu, L.; Chu, Y.; Lukšic, I.; Nair, H.; A McAllister, D.; Campbell, H.; et al.Burden of Streptococcus pneumoniae and Haemophilus influenzae type b disease in children in the era of conjugate vaccines:Global, regional, and national estimates for 2000–15. Lancet Glob. Heal. 2018, 6, e744–e757. [CrossRef]
15. Seale, A.C.; Bianchi-Jassir, F.; Russell, N.J.; Kohli-Lynch, M.; Tann, C.J.; Hall, J.; Madrid, L.; Blencowe, H.; Cousens, S.; Baker,C.J.; et al. Estimates of the Burden of Group B Streptococcal Disease Worldwide for Pregnant Women, Stillbirths, and Children.Clin. Infect. Dis. 2017, 65, S200–S219. [CrossRef]
16. Okike, I.O.; Johnson, A.P.; Henderson, K.L.; Blackburn, R.M.; Muller-Pebody, B.; Ladhani, S.N.; Anthony, M.; Ninis, N.; Heath,P.T. Incidence, Aetiology and Outcome of Bacterial Meningitis in Infants Aged <90 days in the UK and Republic of Ireland:Prospective, enhanced, national population-based surveillance. Clin. Infect. Dis 2014. [CrossRef]
17. Ku, L.C.; Boggess, K.A.; Cohen-Wolkowiez, M. Bacterial Meningitis in Infants. Clin. Perinatol. 2015, 42, 29–45. [CrossRef]18. English, M.; Esamai, F.; Wasunna, A.; Were, F.; Ogutu, B.; Wamae, A.; Snow, R.W.; Peshu, N. Assessment of inpatient paediatric
care in first referral level hospitals in 13 districts in Kenya. Lancet 2004, 363, 1948–1953. [CrossRef]19. Okomo, U.; Dibbasey, T.; Kassama, K.; E Lawn, J.; Zaman, S.M.A.; Kampmann, B.; Howie, S.R.C.; Bojang, K. Neonatal admissions,
quality of care and outcome: 4 years of inpatient audit data from The Gambia’s teaching hospital. Paediatr. Int. Child Heal. 2015,35, 252–264. [CrossRef]
20. United Nations Department of Economic and Social Affairs Population Division. Population Databases. Available online:https://www.un.org/en/development/desa/population/publications/database/index.asp (accessed on 10 June 2019).
21. Thigpen, M.C.; Whitney, C.G.; Messonnier, N.E.; Zell, E.R.; Lynfield, R.; Hadler, J.L.; Harrison, L.H.; Farley, M.M.; Rein-gold, A.; Bennett, N.M.; et al. Bacterial meningitis in the United States, 1998–2007. N. Engl. J. Med. 2011, 364, 2016–2025.[CrossRef] [PubMed]
22. Swartz, M.N. Bacterial meningitis—A view of the past 90 years. N. Engl. J. Med. 2004, 351, 1826–1828. [CrossRef]23. van de Beek, D.; de Gans, J.; Spanjaard, L.; Weisfelt, M.; Reitsma, J.B.; Vermeulen, M. Clinical features and prognostic factors in
adults with bacterial meningitis. N. Engl. J. Med. 2004, 351, 1849–1859. [CrossRef] [PubMed]24. Boisier, P.; Maínassara, H.B.; Sidikou, F.; Djibo, S.; Kairo, K.K.; Chanteau, S. Case-fatality ratio of bacterial meningitis in the
African meningitis belt: We can do better. Vaccine 2007, 25, A24–A29. [CrossRef] [PubMed]25. Dickinson, F.; E Pérez, A. Bacterial Meningitis in children and adolescents: An observational study based on the national
R.F.; Whitney, A.; et al. Etiologies of bacterial meningitis in Bangladesh: Results from a hospital-based study. Am. J. Trop. Med.Hyg. 2009, 81, 475–483. [CrossRef] [PubMed]
27. Department of Evidence, Information and Research (WHO, Geneva) and Maternal Child Epidemiology Estimation (MCEE).MCEE-WHO Methods and Data Sources for Child Causes of Death 2000–2017. Available online: https://www.who.int/healthinfo/global_burden_disease/childcod_methods_2000_2017.pdf?ua=1 (accessed on 4 February 2019).
28. O’Brien, K.L.; Wolfson, L.J.; Watt, J.P.; Henkle, E.; Deloria-Knoll, M.; McCall, N.; Lee, E.; Mulholland, K.; Levine, O.S.; Cherian,T.; et al. Burden of disease caused by Streptococcus pneumoniae in children younger than 5 years: Global estimates. Lancet 2009,374, 893–902. [CrossRef]
29. Benara, S.K.; Singh, P. Validity of causes of infant death by verbal autopsy. Indian J. Pediatr. 1999, 66, 647–650. [CrossRef] [PubMed]30. World Health Organization. A Standard Verbal Autopsy Method for Investigating Causes of Death in Infants and Children.
Available online: https://www.who.int/csr/resources/publications/surveillance/whocdscsrisr994.pdf?ua=1 (accessed on9 September 2019).
31. Butler, D. Verbal autopsy methods questioned. Nature 2010, 467, 1015. [CrossRef]32. Misganaw, A.; Mariam, D.H.; Araya, T.; Anteneh, A. Validity of verbal autopsy method to determine causes of death among
D.; Staedke, S.G. Validity of Verbal Autopsy Procedures for Determining Malaria Deaths in Different Epidemiological Settings inUganda. PLoS ONE 2011, 6, e26892. [CrossRef]
34. Snow, R.W.; Armstrong, J.R.; Forster, D.; Winstanley, M.T.; Marsh, V.M.; Newton, C.R.; Waruiru, C.; Mwangi, I.; Winstanley, P.A.;Marsh, K. Childhood deaths in Africa: Uses and limitations of verbal autopsies. Lancet 1992, 340, 351–355. [CrossRef]
35. Anglewicz, P.; Kohler, H.P. Overestimating HIV infection: The construction and accuracy of subjective probabilities of HIVinfection in rural Malawi. Demogr. Res. 2009, 20, 65–96. [CrossRef] [PubMed]
36. Chandler, C.I.; Jones, C.; Boniface, G.; Juma, K.; Reyburn, H.; Whitty, C.J. Guidelines and mindlines: Why do clinical staffover-diagnose malaria in Tanzania? A qualitative study. Malar. J. 2008, 7, 53. [CrossRef]
37. Gwer, S.; Newton, C.R.; Berkley, J.A. Over-diagnosis and co-morbidity of severe malaria in African children: A guide for clinicians.Am. J. Trop. Med. Hyg. 2007, 77, 6–13. [CrossRef] [PubMed]
38. GBD 2019 Diseases Injuries Collaborators. Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019:A systematic analysis for the Global Burden of Disease Study 2019. Lancet 2020, 396, 1204–1222. [CrossRef]
39. Mathers, C.D. History of global burden of disease assessment at the World Health Organization. Arch. Public Health 2020, 78,1–13. [CrossRef]
40. Meningitis Research Foundation. Meningitis Progress Tracker. Available online: https://public.tableau.com/profile/meningitis.research.foundation#!/vizhome/MeningitisandNeonatalSepsisTracker_15543731247200/CasesandDeaths (accessed on 14September 2020).
41. Traore, Y.; Tameklo, T.A.; Njanpop-Lafourcade, B.-M.; Lourd, M.; Yaro, S.; Niamba, D.; Drabo, A.; E Mueller, J.; Koeck, J.-L.;Gessner, B.D. Incidence, Seasonality, Age Distribution, and Mortality of Pneumococcal Meningitis in Burkina Faso and Togo. Clin.Infect. Dis. 2009, 48, S181–S189. [CrossRef] [PubMed]
42. Palmgren, H. Meningococcal disease and climate. Glob. Health Action 2009, 2. [CrossRef] [PubMed]43. Lingani, C.; Bergeron-Caron, C.; Stuart, J.M.; Fernandez, K.; Djingarey, M.H.; Ronveaux, O.; Schnitzler, J.C.; Perea, W.A.
Meningococcal Meningitis Surveillance in the African Meningitis Belt, 2004–2013. Clin. Infect. Dis. 2015, 61, S410–S415. [CrossRef]44. World Health Organization. Meningitis Weekly Bulletin. Available online: https://www.who.int/emergencies/diseases/
C.; et al. MenAfriNet: A Network Supporting Case-Based Meningitis Surveillance and Vaccine Evaluation in the Meningitis Beltof Africa. J. Infect. Dis. 2019, 220, S148–S154. [CrossRef]
46. Wolfson, L.J.; Strebel, P.M.; Gacic-Dobo, M.; Hoekstra, E.J.; McFarland, J.W.; Hersh, B.S.; Measles, I. Has the 2005 measles mortalityreduction goal been achieved? A natural history modelling study. Lancet 2007, 369, 191–200. [CrossRef]
47. Castillo, P.; Ussene, E.; Ismail, M.R.; Jordao, D.; Lovane, L.; Carrilho, C.; Lorenzoni, C.; Lacerda, M.V.; Palhares, A.; Rodríguez-Carunchio, L.; et al. Pathological Methods Applied to the Investigation of Causes of Death in Developing Countries: MinimallyInvasive Autopsy Approach. PLoS ONE 2015, 10, e0132057. [CrossRef] [PubMed]
48. Bassat, Q.; Ordi, J.; Vila, J.; Ismail, M.R.; Carrilho, C.; Lacerda, M.; Munguambe, K.; Odhiambo, F.; Lell, B.; Sow, S.; et al. Develop-ment of a post-mortem procedure to reduce the uncertainty regarding causes of death in developing countries. Lancet Glob. Health2013, 1, e125–e126. [CrossRef]
49. World Health Organization. World Health Statistics 2017. Monitoring Health for the SDGs. Available online: https://apps.who.int/iris/bitstream/handle/10665/255336/9789241565486-eng.pdf;jsessionid=C00C94A637FA197A27B773D735A711A4?sequence=1 (accessed on 19 September 2019).
50. Thatte, N.; Kalter, H.D.; Baqui, A.H.; Williams, E.M.; Darmstadt, G.L. Ascertaining causes of neonatal deaths using verbal autopsy:Current methods and challenges. J. Perinatol. 2009, 29, 187–194. [CrossRef] [PubMed]
51. GBD 2017 Diseases Injuries Collaborators. Global, regional, and national incidence, prevalence, and years lived with disabilityfor 354 diseases and injuries for 195 countries and territories, 1990–2017: A systematic analysis for the Global Burden of DiseaseStudy 2017. Lancet 2018, 392, 1789–1858. [CrossRef]
